WEE1 inhibition induces anti-tumor immunity by activating ERV and the dsRNA pathway

Abstract

Targeted therapies represent attractive combination partners with immune checkpoint blockade (ICB) to increase the population of patients who benefit or to interdict the emergence of resistance. We demonstrate that targeting WEE1 up-regulates immune signaling through the double-stranded RNA (dsRNA) viral defense pathway with subsequent responsiveness to immune checkpoint blockade even in cGAS/STING-deficient tumors, which is a typical phenotype across multiple cancer types. WEE1 inhibition increases endogenous retroviral elements (ERVs) expression by relieving SETDB1/H3K9me3 repression through down-regulating FOXM1. ERVs trigger dsRNA stress and interferon response, increasing recruitment of anti-tumor T cells with concurrent PD-L1 elevation in multiple tumor models. Furthermore, combining WEE1 inhibition and PD-L1 blockade induced striking tumor regression in a CD8+ T cell-dependent manner. A WEE1 inhibition-induced viral defense signature provides a potentially informative biomarker for patient selection for combination therapy with WEE1 and ICB. WEE1 inhibition stimulates anti-tumor immunity and enhances sensitivity to ICB, providing a rationale for the combination of WEE1 inhibitors and ICB in clinical trials.

Graphical abstract

Figure S1.

AZD1775 activates immunogenic response and IFN pathway. (A) Kaplan–Meier survival curves according to different IFN pathway gene set expressions of ovarian cancer in TCGA. (B) ELISA detection of IFN-β and IFN-λ3 levels in the OV90 and OV2008 cell supernatant treated with AZD1775 or DMSO for 72 h (n = 2; three independent experiments). (C) Heatmap of qPCR quantification of selected genes differentially expressed in ID8 cells treated with DMSO or AZD1775 for 48 h (n = 3). (D) HALLMARK-IFN–enriched pathways were enriched with GSEA analysis in OV90 cells treated with AZD1775 (n = 3). (E) Heatmap of IFN pathway genes analyzed by RNA-seq in OV90 cells after treatment with 800 nM AZD1775 or DMSO for 48 h (n = 3). (F) ELISA detection of CXCL10 levels in OV90 cells treated with AZD1775 or DMSO for 72 h (n = 2; three independent experiments). *, P < 0.05; **, P < 0.01; ***, P < 0.001, as determined by log-rank test (A) and unpaired t test (B and F). P-val, P value; FDR q, false discovery rate p-value.

Figure 1.

AZD1775 activates tumor immune response by IFN pathway and recruiting T cell infiltration. (A) Z-score of relative Irf and NF-κB activity in HCT116-Dual cells after treatment with various agents as indicated (treatment/DMSO) for 48 h. Data represent three independent experiments. (B–D) RNA-seq data analysis of ID8 cells treated with AZD1775 or DMSO (three repeats). (B) Volcano plot of differentially expressed genes. (C) GO and KEGG terms enriched after AZD1775 treatment. (D) GSEA of HALLMARK-IFN pathways induced by AZD1775. (E–I) In vivo experiment of ID8 treated with AZD1775 (n = 5; two independent experiments). Representative images (E) and relative total flux of luminescence (F) of ID8 tumors in C57BL/6 and NOD/SCID mice treated with vehicle or AZD1775. (G) Kaplan–Meier survival curves of C57BL/6 and NOD/SCID mice with ID8 tumors treated as indicated. (H and I) Representative images and quantification of immunohistochemistry analyses of CD3⁺, CD8⁺, CD4⁺, and Foxp3⁺ T reg cells in ID8 tumors after treatments in C57BL/6 mice. HPF, high power field; P-val, P value; FDR q, False Discovery Rate p-value. Scale bar, 50 µm. **, P < 0.01, n.s., not significant as determined by unpaired t test (F and I) and log-rank test (G). p, photons; sr, steradian.

Figure 2.

AZD1775 induced ERVs transcription are responsible for IFN responses in ovarian cancer cells. (A) Western blot of basic expression of dsDNA and dsRNA sensor proteins in 11 ovarian cancer cell lines. Data represent three independent experiments. (B) Heatmaps of differentially expressed ERVs after AZD1775 for 48 h in ID8 cells. (C and D) Quantification of nine randomly selected ERVs by qPCR in ID8 and OV90 cells treated with DMSO or AZD1775 for 48 h (three independent experiments). (E) Cellular dsRNA was evaluated with anti-dsRNA (J2) immunofluorescence in ID8 cells with DMSO or AZD1775 for 72 h. RNase III was used as a negative control for dsRNA signal. Scale bars, 20 µm (three independent experiments). (F) Quantification of ISGs by qPCR in OV90 cells after AZD1775 for 48 h (three independent experiments). (G) MAVS protein polymerization detected by immunoblot using SDD-AGE (Materials and methods) in ID8 and OV90 cells treated with DMSO or AZD1775 for 72 h. Data represent three independent experiments. (H) Heatmap of relative expression of ISGs are shown. OVCAR4 cells were transfected with or without siRNA as indicated and treated with DMSO or AZD1775 for 48 h. Data are representative of three experiments. (I) Heatmap of differentially expressed genes after AZD1775 detected by NanoString immune panel in OVCAR4 cells with siRNA inhibition (n = 1) of the indicated dsDNA/dsRNA sensors. The qPCR data were normalized to β-actin. Data across panels represent mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001, as determined by unpaired t test (C, D, and F).

Figure S2.

WEE1 inhibition induces generation of dsRNA and activation of IFN pathway. (A and B) Quantification of the selected ISGs by qPCR in OV90 cells and ES-2 after stimulation with poly I:C or dsDNA90 (n = 2; three independent experiments). (C) Western blot validation of the effect of WEE1 siRNA knockdown in ID8 and OV90 cells. Data represent three independent experiments. (D) Quantification of ISGs by qPCR in ES-2 cells after AZD1775 treatment for 48 h (three independent experiments). (E and F) Cellular dsRNA was evaluated with anti-dsRNA (J2) immunofluorescence in ID8 cells transfected with either si-NC, si-Wee1, or AZD1775 for 72 h. Scale bar, 50 µm (three independent experiments). (G) Quantification of the selected ISGs by qPCR in OV90 cells transfected with either si-NC or si-WEE1 and treated with AZD1775 for 48 h (three independent experiments). (H) Western blot validation of the effect of dsRNA and dsDNA sensors siRNA knockdown in OVCAR4 cells. Data represent three independent experiments. (I) Quantification of the selected ISGs by qPCR in OVCAR4 cells. Cells were transfected with either si-NC or si-TLR3 and treated with AZD1775 for 48 h (three independent experiments). (J) Representative images of immunofluorescence staining of cGAS in ID8 cells treated with DMSO or AZD1775 for 48 h. Scale bar, 20 µm. Data represent three independent experiments. (K) Quantification of the selected ERV and ISGs by qPCR. OV90 cells were treated with DMSO or AZD1775 or didanosine (DDI) or combination for 48 h (three independent experiments). The real-time qPCR data were normalized to β-actin. Data across panels represent mean ± SEM of three independent experiments. *, P < 0.05; **, P < 0.01; ***, P < 0.001. P values were determined by unpaired t test (A, B, and D), and ANOVA with Bonferroni post hoc test (G, I, and K).

Figure S3.

Down-regulation of SETDB1 increased ISG expression, along with ERVs up-regulation. (A and B) Intron-residing retrotransposons (IR) analysis in RNA-seq data of OV90 cells and ID8 cells (n = 3). (C and D) Whole-genome DNA (C) and LTR methylation levels (D) detected by WGBS in OV90 cells after AZD1775 treatment for 48 h (n = 3). (E) The quantification of global DNA methylation levels of OV90 cells after AZD1775 treatment for 48 h detected by MethylFlash Global DNA Methylation (5-mC) ELISA assay (n = 3; three independent experiments). (F) Western blot of indicated proteins in OV90 cells after SETDB1 silencing with siRNA for 72 h. Data represent three independent experiments. (G and H) Quantification of selected ISGs (G) and transcriptions of ERVs (H) by qPCR in OVCAR4 cells transfected with SETDB1 or scramble (si-NC) siRNA for 72 h (three independent experiments). (I) OV90 cells were treated with AZD1775 or DMSO and harvested at indicated hours. SETDB1 and H3K9me3 were detected by Western blot while ISGs (DHX58, DDX60, IFI44, IFIT1, ISG15, and MX1) were detected by qPCR. Fold changes of SETDB1 or H3K9me3 at each time point were calculated by comparing to DMSO after quantification using Image Lab software 6.0.1. The relative fold changes of ISGs (mean fold changes of DHX58, DDX60, IFI44, IFIT1, ISG15, and MX1) at each time point were also calculated by comparing to DMSO-treated cells. Data represent three independent experiments. The qPCR data were normalized to β-actin. Error bars represent SEM between three replicates. *, P < 0.05; **, P < 0.01; ***, P < 0.001; n.s., not significant, as determined by unpaired t test (E, G, and H).

Figure 3.

Decreased H3K9me3 after WEE1 inhibition contributes to dsRNA stress and promotes IFN activation. (A) Western blot of indicated proteins in OV90, OVCAR4, and KK cells treated with AZD1775 for 72 h. Data represent three independent experiments. (B and C) Quantification of ISGs (B) and ERVs (C) by qPCR in OV90 cells treated as indicated for 48 h (three independent experiments). (D–J) CHIP-seq of H3K9me3 analysis in OV90 cells treated with AZD1775 or DMSO (n = 3). (D–F) The observed and expected (10,000 bootstraps) H3K9me3 peaks located at LTR (D), LTR superfamilies (E), and LTR subfamilies (F) in OV90 cells. (G) Line plot showing the distance distribution from all up-regulated ERVs to H3K9me3 peaks in 60-kb flanking windows (3 kb per bin) in OV90 cells. (H) Cumulative frequencies of distance from up-regulated ERVs to the closest peaks. Dotted line represents the cumulative frequencies of distance from a random selection of stable ERVs to their closest peaks in OV90 cells. (I) Line plot of average H3K9me3 signal at up-regulated ERVs in the gene body plus 5 kb upstream of the transcription start site (TSS) and downstream of the transcription end site (TES) and heatmap showing the H3K9me3 signal for each up-regulated ERV in the gene body plus 5 kb upstream of the TSS and 5 kb downstream of TES in OV90 cells. (J) Density distribution (left) and heatmap (right) of H3K9me3 ChIP-seq signal across the up-regulated ISGs transcription start site (TSS, −5/ + 5 kb). ***, P < 0.001, as determined by unpaired t tests (B and C) and one-sided binomial test (D–F).

Figure 4.

FOXM1 binds to the SETDB1 promoter regulating ERVs transcription. (A) Protein lysates from indicated cells treated with AZD1775 were analyzed by RPPA (n = 2). Heatmap of RPPA data representing “rank-ordered” changes induced by AZD1775. (B) Enrichr analysis of transcription factors associated with genes down-regulated in AZD1775-treated ID8. (C) RPPA profiling in OVCAR4 cells after AZD1775 or FOXM1 silencing with siRNA (n = 2). (D) Western blot of indicated proteins in OVCAR4 cells treated as indicated. The data represent three independent experiments. (E) Heatmap of differentially expressed genes detected by NanoString immune panel in parental or ectopic FOXM1 expressing OVCAR4 cells with or without AZD1775 treatment (n = 3). (F) Screenshot of assay for transposase-accessible chromatin with high throughput sequencing (ATAC-seq), H3K4me3, H3K27me3, and FOXM1 ChIP-seq tracks of SETDB1 in MDA-MB-231 cells. (G) ChIP-qPCR analysis of FOXM1 binding in the SETDB1 promoter region with or without AZD1775 treatment in OV90 cells (n = 3; three independent experiments). Data across panels were mean ± SEM. **, P < 0.01, as determined by unpaired t test.

Figure S4.

Association of FOXM1 with SETDB1 and IFN pathway. (A and B) GSEA analysis of enrichment of HALLMARK E2F PATHWAY in RNA-seq data of ID8 and OV90 cells after treatment with AZD1775 or DMSO for 48 h (n = 3). (C and D) The expression of CDKN1A after treatment with AZD1775 or DMSO for 48 h in RNA-seq data of ID8 and OV90 cells (n = 3). (E) GSEA analysis of IFN pathways after inhibition of FOXM1 by siRNA or inhibitors (FDI-6, NB55), or after FOXM1 overexpression (FOXM1; n = 3). (F–H) Quantification of selected ISGs in OV90 cells treated with FOXM1 inhibition (si-FOXM1 or FDI-6) or overexpression (FOXM1-OE; three independent experiments). (I) Pearson’s correlation of SETDB1 and FOXM1 expression in different cancer types from TCGA. Note that every dot represents one cancer type. (J) Screenshot of FOXM1 ChIP-seq tracks of SETDB1 in ECC-1, GM12878, HeLa, HEK293, MCF7, and SK-N-SH cells. Sample numbers are shown (n = 1 each). The qPCR data were normalized to β-actin. Data across panels represent mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001, as determined by unpaired t test (C, D, and F–H). FPKM, fragments per kilobase of exon model per million mapped fragments; P-val, P value.

Figure 5.

AZD1775 induces PD-L1 expression through IFN signaling. (A and B) Western blots and qPCR of PD-L1 levels in ID8 and OV90 cells treated with a concentration gradient of AZD1775 for 48 h. Data represent three independent experiments. (C and D) Cytofluorimetric histograms of PD-L1 obtained after AZD1775 for 48 h in ID8 and OV90 (three independent experiments). (E and F) Quantification of two CD274 variants by qPCR in ID8 cells and OV90 treated with AZD1775 or DMSO for 48 h. CD274 encoded the canonical full-length PD-L1 while CD274-L2A is the predominant solute PD-L1–encoding variant (three independent experiments). (G and H) Quantification of IFN/IFNR-related genes (G) and IFN-related transcription factors (H) by qPCR in ID8 cells treated with AZD1775 or DMSO for 48 h (three independent experiments). (I) Quantification of selected genes by qPCR in OV90 cells treated with AZD1775, JAK inhibitor (ruxolitinib, 5 µM), or combination for 48 h (three independent experiments). (J) Western blots for STAT1 activation and PD-L1 expression in ID8 cells treated with AZD1775, ruxolitinib, or combination. Data represent three independent experiments. (K and L) ELISA detection of IFN-λ3 levels treated as indicated in ID8 and OV90 cells (three independent experiments). (M) Western blots for PD-L1 treated as indicated. Data represent three independent experiments. The qPCR data were normalized to β-actin. Data across panels represent mean ± SEM. **, P < 0.01; ***, P < 0.001, n.s., not significant as determined by unpaired t test (E–H, K, and L), and ANOVA with Bonferroni post hoc test (A, B, and I).

Figure 6.

AZD1775 combined with ICB represses tumor growth in a CD8⁺ T cell–dependent manner. (A) Representative images and quantification of total flux of luminescence of ID8 tumors in C57BL/6 mice after treatment with vehicle, AZD1775, αPD-L1, or combination of AZD1775 and αPD-L1 (n = 5 or 6 each, two independent experiments). (B) Kaplan–Meier survival curves of C57BL/6 mice with ID8 tumors treated as described in A (n = 5 or 6 each, two independent experiments). (C) The percentage of CD3⁺, CD4⁺, and CD8⁺ T cells in CD45⁺ cells in tumor tissues from different treated groups were quantified via flow cytometry (n = 5 or 6 each). (D) PD-L1 expression determined by flow cytometry on tumor cells (left) and DC cells (right) in tumors treated as indicated (n = 5 or 6 each). (E–H) Representative flow cytometry plots and quantification of IFN-γ, granzyme B, and Ki-67–positive T cells (gated on CD8⁺ cells) in ID8 tumors treated as indicated (n = 5 or 6 each). (I and J) Representative flow cytometry plots (I) and quantification (J) of PD-1–positive T cells (gated on CD8⁺ cells) in ID8 tumors treated as indicated (n = 5 or 6 each). (K) Representative flow cytometry plots (n = 5 or 6 each) of CD4⁺ and CD8⁺ T cells (gated on CD3⁺ T cells) in ID8 tumors treated as indicated. (L) Representative flow cytometry plots of Foxp3⁺ T reg cells (gated on CD4⁺ cells) in ID8 tumors treated as indicated (n = 5 or 6 each). (M) The ratio of CD8⁺T cells to T reg cells is shown (n = 5 or 6 each). The data in C–M are from one representative experiment of two performed with similar results. *, P < 0.05; **, P < 0.01; ***, P < 0.001. P values were determined by log-rank test (B), and ANOVA with Bonferroni post hoc test (A, C, D, H, J, and M). MFI, mean fluorescent intensity.

Figure S5.

Efficacy of AZD1775, αPD-L1, and combination therapy in ID8, MC38, CT26, and B16 models. (A) Plot of alanine aminotransferase (ALT), aspartate aminotransferase (AST), urea nitrogen, creatinine, white blood cell (WBC), red blood cell (RBC), platelet (PLT) count, and hemoglobin (HGB) levels in mice with ID8 tumors treated with indicated treatments (n = 5 or 6). Data are representative of two independent experiments. (B) Percentage of CD45-positive cells in freshly isolated cells of tumor tissues from ID8 mice treated with indicated treatments analyzed by flow cytometry (n = 5 or 6; two independent experiments). (C) CD8⁺ T cells were obtained from human blood peripheral blood mononuclear cells using magnetic beads and labeled with CFSE. Then, CD8⁺ T cells were treated with DMSO or AZD1775 for 4 d. Proliferation activity was detected by flow cytometry with FITC. The data represent three independent experiments. (D) C57BL/6 mice received 200 μg anti-CD8α monoclonal antibody 3 d before MC38 challenge and consolidated on the 0, 3rd, 8th, 14th, 20th, and 26th day after MC38 challenge. These mice were treated from the third day after challenge with indicated treatments. The tumor growth was recorded every 3 d. (n = 5–7). (E) Quantification of IFNs in MC38 tumors treated with vehicle, AZD1775, αPD-L1, or combination (n = 5–7). (F) Heatmap of normalized expression of PD-L1 and selected ERVs versus β-actin in MC38 tumors treated with indicated treatments (n = 5 or 6). (G and H) Tumor growth curves of MAVS and STING-defective MC38 treated with indicated treatments (n = 5). (I and M) Heatmap of normalized expression of PD-L1 and selected ERVs versus β-actin in CT26 (n = 5) and B16 (n = 3) tumors treated with indicated treatments. (J and O) Quantification of selected ERVs (J and N) and ISGs (K and O) expression by qPCR in CT26 and B16 cells treated with AZD1775 or DMSO cells for 48 h in vitro (three independent experiments). (L and P) Western blot of PD-L1 expression in CT26 and B16 cells treated with a series of concentration of AZD1775 or DMSO. Three independent experiments. The qPCR data were normalized to β-actin. Data across panels represent mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001. P values were determined by unpaired t test (J, K, N, and O) and ANOVA with Bonferroni post hoc test (B, D–I, and M).

Figure 7.

AZD1775 combined with ICB assessment in mouse models. (A) Spider plot of MC38 tumor growth treated as indicated. Each line represents one mouse. (B) Tumor growth curves of MC38 tumors after treatment as indicated (n = 5–9). (C) Heatmap of differential expression of immune genes in MC38 tumor tissues from indicated groups (n = 2 each). (D and E) Quantification of selected ERV (D) and ISG (E) expression by qPCR in MC38 cells treated with AZD1775 or DMSO cells in vitro for 48 h (three independent experiments). (F) Western blot of PD-L1 expression in MC38 cells treated with a series of concentration of AZD1775 or DMSO for 72 h in vitro. Data represent three independent experiments. (G and H) Tumor growth curves of CT26 and B16 tumors after indicated treatment (n = 5 or 6). (I) Western blot of basic expression of dsDNA and dsRNA sensor protein in mouse cells. Data represent three independent experiments. (J) Quantification of IFI44 expression by qPCR in mouse cells treated with poly I:C (5 µg/ml) or DMSO cells for 48 h. The qPCR data were normalized to ACTIN. Data across panels represent mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001. P values were determined by unpaired t test (D and E), and ANOVA with Bonferroni post hoc test (B, G, and H).

Figure 8.

FOXM1, SETDB1, and IFN pathway activity may be informative biomarkers for response to AZD1775. (A) Cell viability curves of parental or AZD1775-resistant ID8 (ID8R) cells treated with AZD1775 for 72 h (n = 3; three independent experiments). (B) Heatmap of differentially expressed ISGs in parental and ID8R cells. (C) GSEA of HALLMARK-IFN pathways in ID8R cells. (D) Quantification of selected ISGs by qPCR in ID8 and ID8R cells treated with AZD1775 or DMSO for 48 h (three independent experiments). The qPCR data were normalized to β-actin. Data across panels represent mean ± SEM. (E) Western blot of dsDNA and dsRNA sensors in ID8 and ID8R cells. Data represent three independent experiments. (F and G) Cell viability curves of ID8R cells to AZD1775 after transfection of IFNR siRNA or treated with 5 µM ruxolitinib for 48 h (n = 3; three independent experiments). (H) Heatmap of relative expression of selected ISGs (log2[AZD1775/DMSO]) by qPCR in 44 cancer cells (three independent experiments). Cell lines were classified as nonresponse (non-Resp) and response group (Resp) according to the mean fold-changes of ISGs (cutoff = 2). (I) The FOXM1 and SETDB1 levels in nonresponse and response cells are shown (n = 44). (J) Western blot of basic FOXM1 and SETDB1 expression in ID8, CT26, MC38, and B16 cells. Data represent three independent experiments. (K) Representative immunofluorescence images of AO/PI staining (left) in 3D microfluidic ex vivo culture of PDOS after treatment with AZD1775 or DMSO for 24 h, and IHC staining of FOXM1 and SETDB1 (right) in tumor tissues from the same patients (n = 20). (L) Scatter plot of the correlation between the ratio of live to dead cells measured by AO/PI staining and the IHC score of FOXM1 and SETDB1 in 20 ovarian cancer tissues. AO represents live nucleated cells (green). PI represents dead nucleated cells (red). Scale bar, 50 µm. (M) Unsupervised clustering of AZD1775-sensitive (n = 247) or -resistant (n = 247) cancer cells with FOXM1 pathway genes. (N and O) Comparison of AZD1775 sensitivity (AZD1775 AUC: higher AUC means cells were more resistant to AZD1775) according to FOXM1 pathway score (N) or SETDB1 expression (O). High (n = 247) and low (n = 247). *, P < 0.05; **, P < 0.01; ***, P < 0.001; n.s., not significant, as determined by unpaired t test. P-val, P value; FDR q, false discovery rate p-value.